CN108917761A - A kind of accurate positioning method of unmanned vehicle in underground garage - Google Patents
A kind of accurate positioning method of unmanned vehicle in underground garage Download PDFInfo
- Publication number
- CN108917761A CN108917761A CN201810427773.6A CN201810427773A CN108917761A CN 108917761 A CN108917761 A CN 108917761A CN 201810427773 A CN201810427773 A CN 201810427773A CN 108917761 A CN108917761 A CN 108917761A
- Authority
- CN
- China
- Prior art keywords
- module
- point
- ground
- map
- point cloud
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
- G01C21/206—Instruments for performing navigational calculations specially adapted for indoor navigation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C3/00—Measuring distances in line of sight; Optical rangefinders
Abstract
The invention discloses a kind of accurate positioning method of unmanned vehicle in underground garage, this method includes：Module is divided on laser point cloud ground, character column extracts and the effective point selection module of characteristic matching and non-ground, laser available point merge and down-sampled module, grating map generation and three-dimensional map matching module, initial pose module.The thinking of this method is to reject the dynamic barrier in laser point cloud scene using character column, and selection belongs to the laser available point of this body structure of underground garage, then realizes that vehicle is accurately positioned using map matching technology.This method will be well below conventional method to the sensibility of initial pose, and positioning accuracy with higher is particluarly suitable for containing the scene of a large amount of dynamic barriers in underground garage suitable for positioning and navigation of the unmanned vehicle underground garage.
Description
Technical field
The invention belongs to pilotless automobile research field, in particular to a kind of unmanned vehicle is accurate fixed in underground garage
Position method.
Background technique
In recent years, with the rise of artificial intelligence technology of new generation, weight of the pilotless automobile as artificial intelligence technology
It wants proof of algorithm platform to come into being, and shows the development of blowout, it is public especially to have emerged large quantities of foundation at home
Department, has pushed directly on the fast development of this research field.Pilotless automobile is a kind of special wheeled mobile robot,
It is by the various kinds of sensors entrained by itself, fusion perception road environment automatically obtains vehicle location information, plans roadway
Line controls vehicle movement, to realize the autonomous driving functions of vehicle.
Location technology has in unmanned vehicle research field to play a very important role, and is related to the function such as vehicle perception, planning, control
The accurate realization of energy.Have the characteristics that high-precision, high reliability in view of RTK-GPS, most automatic driving vehicle selections are used
GPS is positioned, however there is shelters such as many tunnels, bridge, higher trees in current urban road,
GPS signal is highly susceptible to interfere in such a case and stop, and vehicle location precision wretched insufficiency is asked to solve this
Topic, it has been proposed that using the combined positioning method of GPS+IMU, but the disadvantage is that hindered in GPS signals such as prolonged tunnels
In the environment of gear, it is easy to fail or vehicle starts in the environment that GPS signal is blocked, can not determine vehicle pose.
For this purpose, researcher proposes to carry out perceptual positioning using the self-contained various kinds of sensors of vehicle, this method is first
Priori map is established, is then positioned using vehicle movement model and inertial navigation device, this method is in no GPS signal
Environment in effect it is obvious, but unique unfortunately this method is easy by noise or dynamic disorder in perception environmental data
The influence of object, to influence positioning accuracy.
Summary of the invention
The purpose of the present invention is to provide a kind of accurate positioning method of unmanned vehicle in underground garage, to solve above-mentioned ask
Topic.
To achieve the above object, the present invention uses following technical scheme：
A kind of accurate positioning method of unmanned vehicle in underground garage, this method includes essence of the unmanned vehicle in underground garage
True positioning system, Precise Position System of the unmanned vehicle in underground garage include ground segmentation module M1, feature extraction with
Characteristic matching and the effective point selection module M2 in non-ground, available point merge to be generated and three with down-sampled module M3, grating map
Tie up map-matching module M4 and initial pose module M5；Divide the extraction of module M1 connection features and characteristic matching, Yi Jifei in ground
The effective point selection module M2 in ground and available point merge and down-sampled module M3, feature extraction and characteristic matching and non-ground
Effective point selection module M2 connection available point, which merges, connect grid with down-sampled module M3 with down-sampled module M3, available point merging
Map is generated to be generated and three-dimensional map matching module with three-dimensional map matching module M4, initial pose module M5 connection grating map
M4；
Ground segmentation module M1 includes ground segmentation module M11, ground point cloud M13 and non-ground points cloud M12；
Feature extraction and characteristic matching and the effective point selection module M2 in non-ground include feature extraction M21, characteristic matching
M22 and characteristics map M23,
It includes that point cloud merges M31, down-sampled M33 and final participation match point cloud that available point, which merges with down-sampled module M3,
M34；
It includes grid map generation module M42 and map-matching module that grating map, which is generated with three-dimensional map matching module M4,
M41, for determining accurate pose of the vehicle in underground garage；
Initial pose module M5 provides the initial bit of vehicle with three-dimensional map matching module M4 for generating for grating map
Appearance；
Specifically include following steps：
Step 1, ground segmentation module M11 is realized using Gaussian process regression algorithm to complete 3 D laser scanning scene
Segmentation, obtain ground point cloud M13 and non-ground points cloud M12, ground point cloud M13 and non-ground points cloud M12 be expressed as PGrd
And PNoGrd；Ground point cloud is the point in scene structure, belongs to available point cloud；
Step 2, the feature of non-ground points cloud is extracted, and is matched with the characteristics map established in advance, scene is selected
In validity feature, and then counter push away obtains non-ground available point cloud, is denoted as PNoGrdValid；
Step 3, the non-ground available point cloud P that module M2 is generatedNoGrdValidThe ground point cloud P generated with module M1GrdIt participates in
Merging module M31, obtain whole available point cloud M32, be expressed as PValid, using down-sampled processing module M33, obtain most
The part available point cloud M34 for participating in map match eventually, is expressed as PLastValid.The module is represented by：
PValid=PGrd∪PNoGrdValid, wherein
PLastValid=Downsample (PValid)；
Step 4, it realizes that grating map generates based on Graph-SLAM technology, is realized by putting the registration Algorithm estimated to face
Map match；Assuming that two point sets areWithPoint cloud and grid in real time are respectively corresponded in the present invention
Occupy point map cloud, Cpq(P, Q) indicates the corresponding relationship between point set P and point set Q, which is realized by nearest neighbor search,
NpAnd NqIndicate the number of two centrostigmas, T=(R, t) is the transformation relation between two point sets, including rotation transformation R and
Shift transformation t meets R ∈ SO (3), t ∈ R3, then have：
Wherein, { pi,qi}∈CpqFor a pair of of corresponding points between two point sets；
Step 5, initial pose module M5 provides initial posture information for map-matching module M41, and matching obtains accurate position
Appearance.
Further, segmentation module M11 in ground is split place to collected laser point cloud using Gauss regression process
Reason, obtain ground point cloud and include metope, pillar, vehicle, pedestrian's barrier non-ground points cloud, wherein ground point cloud can
Pitch angle abundant and roll angle information are provided in position fixing process, belong to available point cloud.
Further, feature extraction M21 function is to extract the character column of non-ground points cloud；Characteristics map M23 is on ground
The character column of the garage inner pillar, metope and the Some vehicles that are marked during figure creation is characterized matching M22 and mentions
For template；Characteristic matching M22 is selection distance measure, and the distance between measures characteristic is estimated using cosine similarity：If two
Character column F1And F2Distance be d (F1,F2), then have
Come whether diacritical point cloud belongs to garage geometry information itself with this, retains if belonging to, otherwise delete, obtain
To scene characteristic M24, it is effective to be back-calculated to obtain non-ground finally according to scene characteristic for the character column for column and metope point cloud
Point cloud M25.
Further, the initial pose module M5 is used to provide sufficiently accurate initial bit for map-matching module M41
Appearance；It is divided into two kinds of situations：First, when vehicle normal operation, use the result of previous frame alignment to position as current vehicle first
Beginning pose, second, when vehicle starts in underground garage or when vehicle location fails, need to redefine the first of vehicle
Beginning pose, at this point, starting global reorientation mechanism, it is ensured that the robustness of vehicle location.
Further, when extracting point cloud feature, it is designed as cylindrical space search, to include by entire column as far as possible
Come in；Search space is fast implemented using nearest neighbor search algorithm, and cylindrical body is using two-dimentional Euclidean distance in conjunction with height threshold
It realizes.
Further, character column extracts process, is handled for non-ground points cloud, and last output is that cylinder is special
Sign；Key point is first looked for, key point is selected according to relevant some conspicuousness features, using the method randomly selected；It finds
After key point, is found out in the cylindrical body centered on key point and owned using the method that nearest neighbor search and height threshold limit
Point establishes local coordinate system using principal component analysis technology based on these points；Then it is coordinately transformed, it will be in cylinder
All points in portion are transformed under laser coordinate by coordinate transform, the rotational invariance of character column are realized with this；It is subsequent logical
It crosses every time along cylindrical body middle shaft rotation fixed angle, calculates separately all the points in cylinder and project to tri- planes of x-y, x-z, y-z
Statistic, the statistic be image higher order statistical square and Shannon entropy, be combined into character column finally by sequencing.
Further, the calculating process of character column is as follows：All the points S in cylinder is calculated separately firstcX-y, x-z,
Projection on y-z plane, projected image are expressed as Ixy(i,j)、Ixz(i,j)、Iyz(i, j) then calculates projected image
Central moment and Shannon entropy, shown in following formula：
Wherein：M and n in formula
It is expressed as the High Order Moment μ of imagemnOrder, e is expressed as the Shannon entropy of image, thus obtains partial cylinder feature F0={ Fxy,
Fxz,Fyz}。
Next, by all the points S inside cylindrical bodycAround z-axis with fixed angle θ rotation, projected image and image are calculated
Central moment and Shannon entropy, obtain this postrotational character column Fk={ Fxy,Fxz,Fyz, and so on, after rotation K times
Terminate；Finally the projected image statistical information being calculated after each rotation is combined according to certain arrangement regulation,
Constitute complete character column vector F=F0∪…∪FK。
After feature calculation finishes, judge whether this feature belongs to this body structure of garage, retains or delete with this
Point cloud inside the cylindrical body；
Its feature selecting strategy is as follows：Assuming that the characteristics map of creation is FMAP={ Fv,Fc,Fw, wherein FvIndicate vehicle
Character column, FcIndicate the feature of garage central post, FwIndicate that the feature of metope obtains in characteristic similarity calculating process
Current signature FxSimilarity relationships D={ d (F between characteristics mapx,Fv),d(Fx,Fc),d(Fx,Fw), as a result, according to such as
Lower formula obtains intracorporal attribute of cylinder：
<l,dmax>=argmaxD
Wherein, l indicates the label of feature, and 1 indicates column, and 2 indicate metope, and 3 indicate vehicle, dmaxIndicate FxWith characteristically
The maximum similarity distance of figure, then final current character column FxAttribute be：
It is unknown characteristics attribute if condition is unsatisfactory for, directly rejects in subsequent processing；In underground garage positioning,
The point for only belonging to this body structure of garage can just be accurately positioned, and vehicle, pedestrian etc. are not belonging to the dynamic disorder of this body structure of garage
Object will cause deviations, for this purpose, we will retain the point cloud on metope and column, delete on other dynamic barriers such as vehicle
Point；Non- ground available point cloud PNoGrdValidFor：
PNoGrdValid=(Pc∪Pw)-(Pv∪Pothers)
Wherein, PcIndicate the point on all columns, PwIndicate the point on all metopes, PvIndicate the point on all vehicles,
PothersIndicate the point on other unknown objects.
Further, coordinate transform is illustrated, FuvwIndicate the local coordinate of all the points building in cylinder, FxyzIt indicates by sitting
Laser coordinate system after mark conversion,Indicates coordinate transformational relation；It is terrible in character column building process
To the rotational invariance of feature, coordinate transform is very crucial, and local coordinate is transformed under world coordinates and calculates by it；It is global
Coordinate is it is known that be vehicle laser sensor coordinate, local coordinate FuvwNeed to calculate confirmation, typical method PCA
(Principal Component Analysis) technology, the characteristics of in view of character column, the coordinate system created is：
Wherein, nsFor key point psNormal vector.
Further, it selects FPFH feature as global registration in reorientation mechanism, enables RANSAC algorithm, find out pair
It should be related to determining coarse positioning transformation, determine whether effectively further according to the detection of scene Geometrical consistency, and so on, to each key
Frame does same processing, obtains one group of initial transformation as a result, selecting optimal finally according to the assessment result of each initial transformation
Transformation is used as initial pose；Wherein valuation functions are：
Wherein,It is a two-valued function, for the number that statistical match is successfully put,It indicates after two frame point clouds pass through initial transformation, the minimum range of corresponding points；Valuation functions are said
It is illustrated when minimum range is less than threshold value, it is believed that otherwise successful match fails, by calculating the match condition of all the points, comparison
Total points obtain matched success rate.
Compared with prior art, the present invention has following technical effect：
It is contour to be not readily susceptible to high building, tunnel compared with common Differential GPS Positioning System in urban area circumstance by the present invention
Big blocking for object and be disturbed.This method can realize the positioning of robust in the environment for completely losing GPS signal, uniquely
It is required that being to generate three-dimensional laser in advance to occupy grating map.
For the present invention compared with other map-matching methods, this method real-time is preferable, and it is accurate fixed to be quickly obtained
Position result.
For the present invention compared with other map-matching methods, this method is lower to the susceptibility of initial pose, can be multiple
In the miscellaneous environment without GPS signal, the realization of robust is accurately positioned.
Detailed description of the invention
Fig. 1 is overall system architecture schematic diagram of the invention.
Fig. 2 is typical search space schematic diagram of the invention.
Fig. 3 is exemplary three-dimensional laser scanning figure of the invention.
Fig. 4 is that character column of the invention extracts flow chart.
Fig. 5 is character column detailed schematic of the invention.
Fig. 6 is coordinate transform schematic diagram of the invention.
Fig. 7 is three kinds of shape schematic diagrames of the invention.
Fig. 8 is the character column schematic diagram of three kinds of shapes of the invention.
Fig. 9 is acquisition initial alignment result schematic diagram of the invention.
Figure 10 is that underground garage three-dimensional of the invention occupies grating map.
Figure 11 is rejecting obstacle object point cloud result figure one of the invention.
Figure 12 is rejecting obstacle object point cloud result figure two of the invention.
Figure 13 is vehicle location result figure of the invention.
Figure 14 is three direction positioning result situations of change of x, y, z of the invention.
Specific embodiment
Below in conjunction with attached drawing, the present invention is described in detail：
Referring to Fig. 1, it is overall system architecture schematic diagram of the invention, including five functional modules, is respectively：
Module M1 is divided on ground, the purpose is to divide module M11 by ground, obtains ground point cloud M13 and non-ground points
Cloud M12, wherein ground segmentation module is realized using Gaussian process regression algorithm.Ground point cloud and non-ground points cloud are expressed as
PGrdAnd PNoGrd；
Feature extraction and characteristic matching and the effective point selection module M2 in non-ground comprising characteristic extracting module M21,
Characteristic matching module M22 obtains scene characteristic M24, then is back-calculated to obtain non-ground available point cloud M25, additionally includes and makes in advance
The characteristics map M23 of work, wherein characteristic extracting module M21 is to extract character column in non-ground available point cloud, and the cylinder is special
Levying has very high efficiency to column, the metope in detection underground garage, and characteristic matching module M22 is according to extracted cylinder
The characteristics of feature, finds suitable distance measure and carries out distance metric to each feature, uses cosine similarity in the present invention
Estimate：If two character column F1And F2Distance be d (F1,F2), then have
Available point merges and down-sampled module M3, the non-ground available point cloud P generated including module M2NoGrdValidWith module
The ground point cloud P that M1 is generatedGrdThe merging module M31 of participation obtains whole available point cloud M32, is expressed as PValid, using
Down-sampled processing module M33 obtains the part available point cloud M34 for finally participating in map match, is expressed as PLastValid.The module
It is represented by：
PValid=PGrd∪PNoGrdValid, wherein
PLastValid=Downsample (PValid)
Grating map, which is generated, generates mould with three-dimensional map matching module M4, including map-matching module M41 and grating map
Block M42, wherein grating map generation module M42 is realized using Graph-SLAM technology, and map-matching module M41 is arrived based on point
The registration Algorithm that face is estimated is realized.Assuming that two point sets areWithIt respectively corresponds in the present invention in real time
Point cloud and grid occupy point map cloud, Cpq(P, Q) indicates the corresponding relationship between point set P and point set Q, which passes through arest neighbors
Search realization, NpAnd NqIndicate the number of two centrostigmas, T=(R, t) is the transformation relation between two point sets, including rotation
R and shift transformation t are changed in transformation, meet R ∈ SO (3), t ∈ R3, then have：
Wherein, { pi,qi}∈CpqFor a pair of of corresponding points between two point sets.
Initial pose module M5 provides initial posture information for map-matching module M41, to accelerate convergence, mentions
High location efficiency, reorientation mechanism are shown in the explanation of Fig. 9.
It referring to fig. 2, is typical search space schematic diagram of the invention.It is common to search for when calculating extraction point cloud feature
Space is cube and spherical shape, and cube is that the basic unit of division three-dimensional space more intuitively facilitates understanding, spherical designs
Interior point will be all considered as within for radius, the present invention is designed as cylindrical body to detect the column and metope in underground garage
Space search, to be as far as possible included entire column.Three of the above search space utilizes nearest neighbor search algorithm fast
Speed realizes that wherein cube uses manhatton distance, and spherical shape uses Euclidean distance, and cylindrical body is using two-dimentional Euclidean distance and height
Threshold value is implemented in combination with.
It is exemplary three-dimensional laser scanning figure of the invention, which is Velodyne company HDL-64E-S3 laser referring to Fig. 3
Sensor laser point cloud collected, scene are Typical Urban crossroad.This sensor effective detection range is up to 100m, energy
Enough obtain the scene three-dimensional information within the scope of 360 degree of vehicle periphery, have the characteristics that interior close outer thin scanning, apart from farther out when error
It is larger.
Referring to fig. 4, process is extracted for character column of the invention, is handled for non-ground points cloud, last output
For character column.First look for key point, it is common practice to select key point, this hair according to relevant some conspicuousness features
It is bright using the method that randomly selects, simple and fast, unfortunately certain points can be calculated twice or repeatedly, by Experimental comparison
It was found that this choosing method influences less time loss.After finding key point, limited using nearest neighbor search and height threshold
The method of system finds out all the points in the cylindrical body centered on key point, based on these points, utilizes principal component analysis technology
(PCA) local coordinate system is established.It is then coordinately transformed, by all points of cylindrical inside by coordinate transform, is transformed into sharp
Under light coordinate, the rotational invariance of character column is realized with this.Subsequent step is used to calculate rotation Statistically invariant feature, leads to
It crosses every time along cylindrical body middle shaft rotation fixed angle, calculates separately all the points in cylinder and project to tri- planes of x-y, x-z, y-z
Statistic, the statistic be image higher order statistical square and Shannon entropy, be combined into character column finally by sequencing.
Referring to Fig. 5, illustrate for character column details of the invention, symbol description in figure：
psFor key point；
poFor cylinder bottom central point；
piFor cylindrical body inside any point；
R is cylinder radius；
Z is cylinder height；
W is the height of x-z, the width of y-z plane projected image and x-y plane projected image；
H is the height of x-z, y-z plane projected image；
dzFor piPoint arrives the distance of cylinder baseplane, meets 0 < dz≤Z；
dxyFor piPoint meets 0 < d in the distance for projecting to bottom centre's point of cylindrical base planexy≤r；
Assuming that ScFor the set that all the points in cylindrical body are constituted, then there is pi∈Sc。
The calculating process of character column is as follows：All the points S in cylinder is calculated separately firstcOn x-y, x-z, y-z plane
Projection, projected image is expressed as Ixy(i,j)、Ixz(i,j)、Iyz(i, j), then calculate projected image central moment and
Shannon entropy, shown in following formula：
Wherein：M and n in formula
It is expressed as the High Order Moment μ of imagemnOrder, e is expressed as the Shannon entropy of image, thus obtains partial cylinder feature F0={ Fxy,
Fxz,Fyz}。
Next, by all the points S inside cylindrical bodycAround z-axis with fixed angle θ rotation, projected image and image are calculated
Central moment and Shannon entropy, obtain this postrotational character column Fk={ Fxy,Fxz,Fyz, and so on, after rotation K times
Terminate.Finally the projected image statistical information being calculated after each rotation is combined according to certain arrangement regulation,
Constitute complete character column vector F=F0∪…∪FK.In method disclosed by the invention, ScAround the z-axis corotating of local coordinate
Three times, each angle is 30 degree, and for more abundant character column, the angle of rotation is suitably adjusted.
It after feature calculation finishes, needs timely to judge whether this feature belongs to this body structure of garage, be protected with this
Stay or delete the point cloud inside the cylindrical body.Its feature selecting strategy is as follows：Assuming that the characteristics map of creation is FMAP={ Fv,
Fc,Fw, wherein FvIndicate the character column of vehicle, FcIndicate the feature of garage central post, FwThe feature for indicating metope, in feature
During Similarity measures, current signature F is obtainedxSimilarity relationships D={ d (F between characteristics mapx,Fv),d(Fx,Fc),
d(Fx,Fw), obtain intracorporal attribute of cylinder according to the following formula as a result,：
<l,dmax>=argmaxD
Wherein, l indicates the label of feature, and 1 indicates column, and 2 indicate metope, and 3 indicate vehicle, dmaxIndicate FxWith characteristically
The maximum similarity distance of figure, then final current character column FxAttribute be：
It is unknown characteristics attribute if condition is unsatisfactory for, directly rejects in subsequent processing.It is fixed in underground garage
In position, the point for only belonging to this body structure of garage can be just accurately positioned, and vehicle, pedestrian etc. are not belonging to the dynamic of this body structure of garage
Barrier will cause deviations, for this purpose, we will retain the point cloud on metope and column, delete other dynamic disorders such as vehicle
Point on object.Non- ground available point cloud PNoGrdValidFor：
PNoGrdValid=(Pc∪Pw)-(Pv∪Pothers)
Wherein, PcIndicate the point on all columns, PwIndicate the point on all metopes, PvIndicate the point on all vehicles,
PothersIndicate the point on other unknown objects.
Referring to Fig. 6, illustrate for coordinate transform of the invention, FuvwIndicate the local coordinate of all the points building in cylinder, Fxyz
Indicate the laser coordinate system after coordinate is converted,Indicates coordinate transformational relation.In character column building process
In, the rotational invariance of feature in order to obtain, coordinate transform is very crucial, and local coordinate is transformed under world coordinates and carries out by it
It calculates.World coordinates in the present invention is it is known that be vehicle laser sensor coordinate, local coordinate FuvwNeed to calculate confirmation, typically
Method is PCA (Principal Component Analysis) technology, the characteristics of in view of character column, the coordinate system that is created
For：
Wherein, nsFor key point psNormal vector.
Referring to Fig. 7, for the cylinder spy that three kinds of typical shape models and Fig. 8 of the invention are three kinds of typical shape models
Sign.This model is just for the structure in common underground garage, it is generally recognized that and it include metope, column, vehicle, it is seen from figure 7 that
The character column of three kinds of typical shape models is calculated under global coordinate system, is that the cylinder of correspondingly-shaped in Fig. 7 is special in Fig. 8
Sign, sees from figure, and character column has very strong distinguishing ability, and column and metope are separated from complicated scene.
Referring to Fig. 9, illustrate for acquisition initial alignment result of the invention, this figure is described when unmanned vehicle is in underground garage
When starting, or in fixation and recognition, start reorientation mechanism.According to diagram, key frame is point cloud when creating map in figure
Frame, it includes scene point cloud posture information corresponding with its.Inputting information is to put cloud and map key frame in real time, in the present invention
In the key frame be selected as 7 frames, determined according to the range that scene size and every frame point cloud are covered, selection it is more, obtain
The initialization pose arrived can be more accurate, and the corresponding calculating time can be elongated.Select FPFH feature as complete in reorientation mechanism
Office's matching, enables RANSAC algorithm, finds out corresponding relationship and determines that coarse positioning converts, detects and determine further according to scene Geometrical consistency
Whether effectively.And so on, same processing is done to each key frame, obtains one group of initial transformation as a result, last according to each
The assessment result of initial transformation selects optimal transformation as initial pose.Wherein valuation functions are：
Wherein,It is a two-valued function, for the number that statistical match is successfully put,It indicates after two frame point clouds pass through initial transformation, the minimum range of corresponding points.Valuation functions are said
It is illustrated when minimum range is less than threshold value, it is believed that otherwise successful match fails, by calculating the match condition of all the points, comparison
Total points obtain matched success rate, the experiment proves that the method effectively assesses the effect of initial transformation.
Referring to Figure 10, grating map is occupied for underground garage three-dimensional of the invention, which passes through Graph-SLAM technology
It realizes.
See that the present invention passes through from figure for the result figure of rejecting obstacle object point cloud of the invention referring to Figure 11 and Figure 12
The method for extracting character column rejects dynamic barrier, retains the point that can be improved scene of positioning accuracy result itself.
Referring to Figure 13 and Figure 14, three direction positioning results of vehicle location result figure respectively of the invention and x, y, z become
Change situation, see that vehicle location track is smooth accurate from Figure 13, in conjunction with Figure 14, the pose that the direction x and y is converts non-ordinary light
Sliding, only there are some fluctuations on the direction z, but fluctuation range meets positioning requirements within 0.1m.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
A specific embodiment of the invention is only limitted to this, for those of ordinary skill in the art to which the present invention belongs, is not taking off
Under the premise of from present inventive concept, several simple deduction or replace are also made, all shall be regarded as belonging to the present invention by being submitted
Claims determine scope of patent protection.
Claims (9)
1. a kind of accurate positioning method of unmanned vehicle in underground garage, which is characterized in that this method is based on unmanned vehicle in underground
Precise Position System in garage, Precise Position System of the unmanned vehicle in underground garage include ground segmentation module M1,
Feature extraction and characteristic matching and the effective point selection module M2 in non-ground, available point merge and down-sampled module M3, grid
Figure generates and three-dimensional map matching module M4 and initial pose module M5；Divide the extraction of module M1 connection features and feature in ground
Match and the effective point selection module M2 in non-ground and available point merge with down-sampled module M3, feature extraction and characteristic matching, with
And the non-effective point selection module M2 connection available point in ground merges and down-sampled module M3, available point merging and down-sampled module M3
Grating map is connected to generate and three-dimensional map matching module M4, initial pose module M5 connection grating map generation and three-dimensional map
Matching module M4；
Ground segmentation module M1 includes ground segmentation module M11, ground point cloud M13 and non-ground points cloud M12；
Feature extraction and characteristic matching and the effective point selection module M2 in non-ground include feature extraction M21, characteristic matching M22
With characteristics map M23,
It includes that point cloud merges M31, down-sampled M33 and final participation match point cloud M34 that available point, which merges with down-sampled module M3,；
It includes grid map generation module M42 and map-matching module M41 that grating map, which is generated with three-dimensional map matching module M4,
For determining accurate pose of the vehicle in underground garage；
Initial pose module M5 provides the initial pose of vehicle with three-dimensional map matching module M4 for generating for grating map；
Specifically include following steps：
Step 1, ground segmentation module M11 divides complete 3 D laser scanning scene using the realization of Gaussian process regression algorithm
It cuts, obtains ground point cloud M13 and non-ground points cloud M12, ground point cloud M13 and non-ground points cloud M12 is expressed as PGrdWith
PNoGrd；Ground point cloud is the point in scene structure, belongs to available point cloud；
Step 2, the feature of non-ground points cloud is extracted, and is matched with the characteristics map established in advance, is selected in scene
Validity feature, and then counter push away obtains non-ground available point cloud, is denoted as PNoGrdValid；
Step 3, the non-ground available point cloud P that module M2 is generatedNoGrdValidThe ground point cloud P generated with module M1GrdThe conjunction of participation
And module M31, whole available point cloud M32 is obtained, P is expressed asValid, using down-sampled processing module M33, finally joined
With the part available point cloud M34 of map match, it is expressed as PLastValid；The module is expressed as：
PValid=PGrd∪PNoGrdValid, wherein
PLastValid=Downsample (PValid)；
Step 4, it realizes that grating map generates based on Graph-SLAM technology, realizes map by putting the registration Algorithm estimated to face
Matching；Assuming that two point sets areWithPoint map cloud is occupied respectively corresponding point cloud and grid in real time,
Cpq(P, Q) indicates the corresponding relationship between point set P and point set Q, which is realized by nearest neighbor search, NpAnd NqIndicate two
The number of point centrostigma, T=(R, t) are the transformation relation between two point sets, including rotation transformation R and shift transformation t, are met
R∈SO(3),t∈R3, then have：
Wherein, { pi,qi}∈CpqFor a pair of of corresponding points between two point sets；
Step 5, initial pose module M5 provides initial posture information for map-matching module M41, and matching obtains accurate pose.
2. a kind of accurate positioning method of the unmanned vehicle according to claim 1 in underground garage, which is characterized in that ground
Divide module M11 and using Gauss regression process processing be split to collected laser point cloud, obtain ground point cloud and comprising
There is the non-ground points cloud of metope, pillar, vehicle, pedestrian's barrier, wherein ground point cloud can provide abundant in position fixing process
Pitch angle and roll angle information, belong to available point cloud.
3. a kind of accurate positioning method of the unmanned vehicle according to claim 1 in underground garage, which is characterized in that feature
Extracting M21 function is to extract the character column of non-ground points cloud；Characteristics map M23 is the vehicle marked during map building
The character column of library inner pillar, metope and Some vehicles is characterized matching M22 and provides template；Characteristic matching M22 is choosing
The distance between distance measure, measures characteristic are selected, is estimated using cosine similarity：If two character column F1And F2Distance be d
(F1,F2), then have
Come whether diacritical point cloud belongs to garage geometry information itself with this, retains if belonging to, otherwise delete, must show up
Scape feature M24 is back-calculated to obtain non-ground available point cloud finally according to scene characteristic for the character column of column and metope point cloud
M25。
4. a kind of accurate positioning method of the unmanned vehicle according to claim 1 in underground garage, which is characterized in that described
Initial pose module M5 is used to provide sufficiently accurate initial pose for map-matching module M41；It is divided into two kinds of situations：First,
When vehicle normal operation, the initial pose for using the result of previous frame alignment to position as current vehicle, second, when vehicle is on ground
When starting in lower garage or when vehicle location fails, need to redefine the initial pose of vehicle, at this point, the global weight of starting
Location mechanism, it is ensured that the robustness of vehicle location.
5. a kind of accurate positioning method of the unmanned vehicle according to claim 1 in underground garage, which is characterized in that extract
When point cloud feature, it is designed as cylindrical space search, to be as far as possible included entire column；Search space is using recently
Adjacent searching algorithm fast implements, and cylindrical body is implemented in combination with using two-dimentional Euclidean distance with height threshold.
6. a kind of accurate positioning method of the unmanned vehicle according to claim 5 in underground garage, which is characterized in that cylinder
Feature extraction process is handled for non-ground points cloud, and last output is character column；Key point is first looked for, according to
Relevant some conspicuousness features select key point, using the method randomly selected；After finding key point, searched using arest neighbors
Rope and the method for height threshold limitation find out all the points in the cylindrical body centered on key point, based on these points, utilize
Principal component analysis technology establishes local coordinate system；It is then coordinately transformed, all points of cylindrical inside is passed through into coordinate transform,
It is transformed under laser coordinate, the rotational invariance of character column is realized with this；Subsequently through every time along cylindrical body middle shaft rotation
Fixed angle calculates separately the statistic that all the points in cylinder project to tri- planes of x-y, x-z, y-z, which is image
Higher order statistical square and Shannon entropy, be combined into character column finally by sequencing.
7. a kind of accurate positioning method of the unmanned vehicle according to claim 6 in underground garage, which is characterized in that cylinder
The calculating process of feature is as follows：All the points S in cylinder is calculated separately firstcProjection on x-y, x-z, y-z plane, perspective view
As being expressed as Ixy(i,j)、Ixz(i,j)、Iyz(i, j) then calculates the central moment and Shannon entropy of projected image, following public
Shown in formula：
Wherein：M and n is indicated in formula
For the High Order Moment μ of imagemnOrder, e is expressed as the Shannon entropy of image, thus obtains partial cylinder feature F0={ Fxy,Fxz,
Fyz}；
By all the points S inside cylindrical bodycAround z-axis with fixed angle θ rotation, the central moment and perfume (or spice) of projected image and image are calculated
Agriculture entropy obtains this postrotational character column Fk={ Fxy,Fxz,Fyz, and so on, terminate after rotation K times；Finally will
The projected image statistical information being calculated after rotation every time is combined according to certain arrangement regulation, constitutes complete circle
Column feature vector F=F0∪…∪FK；
After feature calculation finishes, judge whether this feature belongs to this body structure of garage, retains or delete the circle with this
The point cloud of column body；
Its feature selecting strategy is as follows：Assuming that the characteristics map of creation is FMAP={ Fv,Fc,Fw, wherein FvIndicate the cylinder of vehicle
Feature, FcIndicate the feature of garage central post, FwIndicate that the feature of metope obtains current spy in characteristic similarity calculating process
Levy FxSimilarity relationships D={ d (F between characteristics mapx,Fv),d(Fx,Fc),d(Fx,Fw), as a result, according to the following formula
Obtain intracorporal attribute of cylinder：
<l,dmax>=argmaxD
Wherein, l indicates the label of feature, and 1 indicates column, and 2 indicate metope, and 3 indicate vehicle, dmaxIndicate FxMost with characteristics map
Big similarity distance, then final current character column FxAttribute be：
It is unknown characteristics attribute if condition is unsatisfactory for, directly rejects in subsequent processing；In underground garage positioning, only
The point for belonging to this body structure of garage can just be accurately positioned, and vehicle, pedestrian etc. are not belonging to the dynamic barrier meeting of this body structure of garage
Deviations are caused, for this purpose, we will retain the point cloud on metope and column, are deleted on other dynamic barriers such as vehicle
Point；Non- ground available point cloud PNoGrdValidFor：
PNoGrdValid=(Pc∪Pw)-(Pv∪Pothers)
Wherein, PcIndicate the point on all columns, PwIndicate the point on all metopes, PvIndicate the point on all vehicles, Pothers
Indicate the point on other unknown objects.
8. a kind of accurate positioning method of the unmanned vehicle according to claim 6 in underground garage, which is characterized in that coordinate
Transformation signal, FuvwIndicate the local coordinate of all the points building in cylinder, FxyzIndicate the laser coordinate system after coordinate is converted,Indicates coordinate transformational relation；In character column building process, the rotational invariance of feature, is sat in order to obtain
Mark transformation is very crucial, and local coordinate is transformed under world coordinates and calculates by it；World coordinates is it is known that be vehicle laser
Sensor coordinate, local coordinate FuvwNeed to calculate confirmation, typical method is PCA (Principal Component
Analysis) technology, the characteristics of in view of character column, the coordinate system created is：
Wherein, nsFor key point psNormal vector.
9. a kind of accurate positioning method of the unmanned vehicle according to claim 4 in underground garage, which is characterized in that in weight
It selects FPFH feature as global registration in location mechanism, enables RANSAC algorithm, find out corresponding relationship and determine that coarse positioning converts,
Determine whether effectively further according to the detection of scene Geometrical consistency, and so on, same processing is done to each key frame, obtains one
Group initial transformation selects optimal transformation as initial pose as a result, finally according to the assessment result of each initial transformation；Wherein
Valuation functions are：
Wherein,It is a two-valued function, for the number that statistical match is successfully put,It indicates after two frame point clouds pass through initial transformation, the minimum range of corresponding points；Valuation functions are said
It is illustrated when minimum range is less than threshold value, it is believed that otherwise successful match fails, by calculating the match condition of all the points, comparison
Total points obtain matched success rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810427773.6A CN108917761B (en) | 2018-05-07 | 2018-05-07 | Accurate positioning method of unmanned vehicle in underground garage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810427773.6A CN108917761B (en) | 2018-05-07 | 2018-05-07 | Accurate positioning method of unmanned vehicle in underground garage |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108917761A true CN108917761A (en) | 2018-11-30 |
| CN108917761B CN108917761B (en) | 2021-01-19 |
Family
ID=64403608
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810427773.6A Active CN108917761B (en) | 2018-05-07 | 2018-05-07 | Accurate positioning method of unmanned vehicle in underground garage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108917761B (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109889977A (en) * | 2019-02-25 | 2019-06-14 | 广州市香港科大霍英东研究院 | A kind of bluetooth localization method, device, equipment and system returned based on Gauss |
| CN109903383A (en) * | 2019-04-11 | 2019-06-18 | 中国矿业大学 | A kind of coalcutter accurate positioning method in the threedimensional model of working face coal seam |
| CN110031825A (en) * | 2019-04-17 | 2019-07-19 | 北京智行者科技有限公司 | Laser positioning initial method |
| CN110349192A (en) * | 2019-06-10 | 2019-10-18 | 西安交通大学 | A kind of tracking of the online Target Tracking System based on three-dimensional laser point cloud |
| CN111239763A (en) * | 2020-03-06 | 2020-06-05 | 广州视源电子科技股份有限公司 | Object positioning method and device, storage medium and processor |
| CN111323027A (en) * | 2018-12-17 | 2020-06-23 | 兰州大学 | Method and device for manufacturing high-precision map based on fusion of laser radar and panoramic camera |
| CN111469781A (en) * | 2019-01-24 | 2020-07-31 | 北京京东尚科信息技术有限公司 | Method and apparatus for outputting information |
| CN111968179A (en) * | 2020-08-13 | 2020-11-20 | 厦门大学 | Method for positioning automatic driving vehicle in underground parking lot |
| CN112017219A (en) * | 2020-03-17 | 2020-12-01 | 湖北亿咖通科技有限公司 | Laser point cloud registration method |
| CN112329749A (en) * | 2021-01-05 | 2021-02-05 | 新石器慧通（北京）科技有限公司 | Point cloud labeling method and labeling equipment |
| CN112382116A (en) * | 2020-11-12 | 2021-02-19 | 浙江吉利控股集团有限公司 | Method and system for acquiring point cloud map of vehicle |
| JP2021051057A (en) * | 2019-09-24 | 2021-04-01 | ベイジン バイドゥ ネットコム サイエンス アンド テクノロジー カンパニー リミテッド | Method and apparatus for detecting ground point cloud point |
| CN112700479A (en) * | 2020-12-23 | 2021-04-23 | 北京超星未来科技有限公司 | Registration method based on CNN point cloud target detection |
| CN112710318A (en) * | 2020-12-14 | 2021-04-27 | 深圳市商汤科技有限公司 | Map generation method, route planning method, electronic device, and storage medium |
| CN112923933A (en) * | 2019-12-06 | 2021-06-08 | 北理慧动（常熟）车辆科技有限公司 | Laser radar SLAM algorithm and inertial navigation fusion positioning method |
| CN112991440A (en) * | 2019-12-12 | 2021-06-18 | 纳恩博（北京）科技有限公司 | Vehicle positioning method and device, storage medium and electronic device |
| CN113465607A (en) * | 2021-06-30 | 2021-10-01 | 上海西井信息科技有限公司 | Port vehicle positioning method, port vehicle positioning device, electronic equipment and storage medium |
| CN113568003A (en) * | 2021-07-26 | 2021-10-29 | 奥特酷智能科技(南京)有限公司 | Anti-collision early warning system and method for airport ground service vehicle |
| CN113593021A (en) * | 2021-06-22 | 2021-11-02 | 天津大学 | Garage point cloud map construction method based on contour segmentation |
| CN113778077A (en) * | 2021-02-09 | 2021-12-10 | 贵州京邦达供应链科技有限公司 | Positioning method and device of mobile platform and storage medium |
| CN114280583A (en) * | 2022-03-02 | 2022-04-05 | 武汉理工大学 | Laser radar positioning precision verification method and system under condition of no GPS signal |
| WO2022087916A1 (en) * | 2020-10-28 | 2022-05-05 | 华为技术有限公司 | Positioning method and apparatus, and electronic device and storage medium |
| CN114577215A (en) * | 2022-03-10 | 2022-06-03 | 山东新一代信息产业技术研究院有限公司 | Method, device and medium for updating feature map of mobile robot |
| CN116719067A (en) * | 2023-08-08 | 2023-09-08 | 科沃斯家用机器人有限公司 | Method and apparatus for detecting reference station position variation, and readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103389103A (en) * | 2013-07-03 | 2013-11-13 | 北京理工大学 | Geographical environmental characteristic map construction and navigation method based on data mining |
| KR20150096275A (en) * | 2014-02-14 | 2015-08-24 | 삼성전자주식회사 | Method and device for acquiring information |
| CN104897161A (en) * | 2015-06-02 | 2015-09-09 | 武汉大学 | Indoor planimetric map making method based on laser ranging |
| CN106896353A (en) * | 2017-03-21 | 2017-06-27 | 同济大学 | A kind of unmanned vehicle crossing detection method based on three-dimensional laser radar |
-
2018
- 2018-05-07 CN CN201810427773.6A patent/CN108917761B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103389103A (en) * | 2013-07-03 | 2013-11-13 | 北京理工大学 | Geographical environmental characteristic map construction and navigation method based on data mining |
| KR20150096275A (en) * | 2014-02-14 | 2015-08-24 | 삼성전자주식회사 | Method and device for acquiring information |
| CN104897161A (en) * | 2015-06-02 | 2015-09-09 | 武汉大学 | Indoor planimetric map making method based on laser ranging |
| CN106896353A (en) * | 2017-03-21 | 2017-06-27 | 同济大学 | A kind of unmanned vehicle crossing detection method based on three-dimensional laser radar |
Non-Patent Citations (2)
| Title |
|---|
| 李宁等: "基于多线激光雷达的非结构化道路感知技术研究", 《车辆与动力技术》 * |
| 杨飞等: "基于三维激光雷达的动态障碍实时检测与跟踪", 《浙江大学学报（工学版）》 * |
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111323027A (en) * | 2018-12-17 | 2020-06-23 | 兰州大学 | Method and device for manufacturing high-precision map based on fusion of laser radar and panoramic camera |
| CN111469781A (en) * | 2019-01-24 | 2020-07-31 | 北京京东尚科信息技术有限公司 | Method and apparatus for outputting information |
| CN109889977A (en) * | 2019-02-25 | 2019-06-14 | 广州市香港科大霍英东研究院 | A kind of bluetooth localization method, device, equipment and system returned based on Gauss |
| CN109889977B (en) * | 2019-02-25 | 2021-01-12 | 广州市香港科大霍英东研究院 | Bluetooth positioning method, device, equipment and system based on Gaussian regression |
| CN109903383A (en) * | 2019-04-11 | 2019-06-18 | 中国矿业大学 | A kind of coalcutter accurate positioning method in the threedimensional model of working face coal seam |
| CN110031825A (en) * | 2019-04-17 | 2019-07-19 | 北京智行者科技有限公司 | Laser positioning initial method |
| CN110349192B (en) * | 2019-06-10 | 2021-07-13 | 西安交通大学 | Tracking method of online target tracking system based on three-dimensional laser point cloud |
| CN110349192A (en) * | 2019-06-10 | 2019-10-18 | 西安交通大学 | A kind of tracking of the online Target Tracking System based on three-dimensional laser point cloud |
| JP2021051057A (en) * | 2019-09-24 | 2021-04-01 | ベイジン バイドゥ ネットコム サイエンス アンド テクノロジー カンパニー リミテッド | Method and apparatus for detecting ground point cloud point |
| US11328429B2 (en) | 2019-09-24 | 2022-05-10 | Apollo Intelligent Driving Technology (Beijing) Co., Ltd. | Method and apparatus for detecting ground point cloud points |
| CN112923933A (en) * | 2019-12-06 | 2021-06-08 | 北理慧动（常熟）车辆科技有限公司 | Laser radar SLAM algorithm and inertial navigation fusion positioning method |
| CN112991440B (en) * | 2019-12-12 | 2024-04-12 | 纳恩博（北京）科技有限公司 | Positioning method and device for vehicle, storage medium and electronic device |
| CN112991440A (en) * | 2019-12-12 | 2021-06-18 | 纳恩博（北京）科技有限公司 | Vehicle positioning method and device, storage medium and electronic device |
| CN111239763A (en) * | 2020-03-06 | 2020-06-05 | 广州视源电子科技股份有限公司 | Object positioning method and device, storage medium and processor |
| CN112017219B (en) * | 2020-03-17 | 2022-04-19 | 湖北亿咖通科技有限公司 | Laser point cloud registration method |
| CN112017219A (en) * | 2020-03-17 | 2020-12-01 | 湖北亿咖通科技有限公司 | Laser point cloud registration method |
| CN111968179A (en) * | 2020-08-13 | 2020-11-20 | 厦门大学 | Method for positioning automatic driving vehicle in underground parking lot |
| CN111968179B (en) * | 2020-08-13 | 2022-07-19 | 厦门大学 | Automatic driving vehicle positioning method for underground parking garage |
| WO2022087916A1 (en) * | 2020-10-28 | 2022-05-05 | 华为技术有限公司 | Positioning method and apparatus, and electronic device and storage medium |
| CN112382116A (en) * | 2020-11-12 | 2021-02-19 | 浙江吉利控股集团有限公司 | Method and system for acquiring point cloud map of vehicle |
| CN112710318A (en) * | 2020-12-14 | 2021-04-27 | 深圳市商汤科技有限公司 | Map generation method, route planning method, electronic device, and storage medium |
| CN112700479B (en) * | 2020-12-23 | 2024-02-23 | 北京超星未来科技有限公司 | Registration method based on CNN point cloud target detection |
| CN112700479A (en) * | 2020-12-23 | 2021-04-23 | 北京超星未来科技有限公司 | Registration method based on CNN point cloud target detection |
| CN112329749B (en) * | 2021-01-05 | 2021-04-27 | 新石器慧通（北京）科技有限公司 | Point cloud labeling method and labeling equipment |
| CN112329749A (en) * | 2021-01-05 | 2021-02-05 | 新石器慧通（北京）科技有限公司 | Point cloud labeling method and labeling equipment |
| CN113778077A (en) * | 2021-02-09 | 2021-12-10 | 贵州京邦达供应链科技有限公司 | Positioning method and device of mobile platform and storage medium |
| CN113778077B (en) * | 2021-02-09 | 2024-04-16 | 贵州京邦达供应链科技有限公司 | Positioning method and equipment for mobile platform and storage medium |
| CN113593021A (en) * | 2021-06-22 | 2021-11-02 | 天津大学 | Garage point cloud map construction method based on contour segmentation |
| CN113593021B (en) * | 2021-06-22 | 2023-06-09 | 天津大学 | Garage point cloud map construction method based on contour segmentation |
| CN113465607A (en) * | 2021-06-30 | 2021-10-01 | 上海西井信息科技有限公司 | Port vehicle positioning method, port vehicle positioning device, electronic equipment and storage medium |
| CN113568003A (en) * | 2021-07-26 | 2021-10-29 | 奥特酷智能科技(南京)有限公司 | Anti-collision early warning system and method for airport ground service vehicle |
| CN113568003B (en) * | 2021-07-26 | 2022-11-01 | 奥特酷智能科技(南京)有限公司 | Anti-collision early warning system and method for airport ground service vehicle |
| CN114280583A (en) * | 2022-03-02 | 2022-04-05 | 武汉理工大学 | Laser radar positioning precision verification method and system under condition of no GPS signal |
| CN114577215A (en) * | 2022-03-10 | 2022-06-03 | 山东新一代信息产业技术研究院有限公司 | Method, device and medium for updating feature map of mobile robot |
| CN114577215B (en) * | 2022-03-10 | 2023-10-27 | 山东新一代信息产业技术研究院有限公司 | Method, equipment and medium for updating characteristic map of mobile robot |
| CN116719067B (en) * | 2023-08-08 | 2023-10-17 | 科沃斯家用机器人有限公司 | Method and apparatus for detecting reference station position variation, and readable storage medium |
| CN116719067A (en) * | 2023-08-08 | 2023-09-08 | 科沃斯家用机器人有限公司 | Method and apparatus for detecting reference station position variation, and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108917761B (en) | 2021-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108917761A (en) | A kind of accurate positioning method of unmanned vehicle in underground garage | |
| Li et al. | Springrobot: A prototype autonomous vehicle and its algorithms for lane detection | |
| Wang et al. | Torontocity: Seeing the world with a million eyes | |
| Rozsa et al. | Obstacle prediction for automated guided vehicles based on point clouds measured by a tilted LIDAR sensor | |
| Holgado‐Barco et al. | Automatic inventory of road cross‐sections from mobile laser scanning system | |
| Yu et al. | Automated extraction of urban road facilities using mobile laser scanning data | |
| CN109631855A (en) | High-precision vehicle positioning method based on ORB-SLAM | |
| US20190137280A1 (en) | System and method for precision localization and mapping | |
| Wang et al. | Intelligent vehicle self-localization based on double-layer features and multilayer LIDAR | |
| CN103278170A (en) | Mobile robot cascading map building method based on remarkable scenic spot detection | |
| CN103247040A (en) | Layered topological structure based map splicing method for multi-robot system | |
| CN115388902B (en) | Indoor positioning method and system, AR indoor positioning navigation method and system | |
| Nagy et al. | 3D CNN-based semantic labeling approach for mobile laser scanning data | |
| Zhang et al. | An efficient LiDAR-based localization method for self-driving cars in dynamic environments | |
| Liu et al. | Deep-learning and depth-map based approach for detection and 3-D localization of small traffic signs | |
| CN115564865A (en) | Construction method and system of crowdsourcing high-precision map, electronic equipment and vehicle | |
| Li et al. | Visual map-based localization for intelligent vehicles from multi-view site matching | |
| Liu et al. | Dloam: Real-time and robust lidar slam system based on cnn in dynamic urban environments | |
| Wu et al. | A stepwise minimum spanning tree matching method for registering vehicle-borne and backpack LiDAR point clouds | |
| Zhou et al. | Asl-slam: A lidar slam with activity semantics-based loop closure | |
| Yuan et al. | A novel approach to image-sequence-based mobile robot place recognition | |
| Lu et al. | Pole-based localization for autonomous vehicles in urban scenarios using local grid map-based method | |
| Wang et al. | Pole-like objects mapping and long-term robot localization in dynamic urban scenarios | |
| Shi et al. | A fast LiDAR place recognition and localization method by fusing local and global search | |
| Zhou et al. | Place recognition and navigation of outdoor mobile robots based on random Forest learning with a 3D LiDAR |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |